Abstract
The recent article 1 on the diagnosis and investigation of adrenal insufficiency in adults was comprehensive. I wish however to make several comments.
Firstly, adrenal insufficiency caused by pituitary/hypothalamic disease (non-iatrogenic) is almost always associated with impediments in other pituitary axes. Furthermore, these axes are usually affected earlier than the hypothalamic–pituitary–adrenal (HPA) axis. Adrenal insufficiency caused by pituitary or hypothalamic disease should not be regarded nor investigated as an entity in its own right, with the exception of isolated adrenocorticotrophin deficiency, which is rare. Initial laboratory investigation should focus on other hormones. Tests such as serum testosterone in men, serum gonadotrophins in postmenopausal women or women with amenorrhoea/irregular periods and serum prolactin are informative and, in addition to thyroid function tests, are important. Initial investigation should therefore include these readily available inexpensive tests. Interpreted in the context of other clinical correlates quoted by the authors, diagnosis can be surmised, followed by dedicated HPA-axis investigation as described 1 if deemed necessary. Over the last two decades or so investigations such as the insulin tolerance test, metyrapone and glucagon stimulation tests have become obsolete or of very limited clinical use.
I am also concerned about the use of the term ‘diagnostic cut-off levels’ in relation to serum cortisol concentration in various situations. The use of this term is misleading. At the cut-off levels indicated diagnosis cannot be made since the odds ratio at the cut-off is equal to 1 (i.e. an odds ratio of 1 indicates that the condition under study is equally likely in both groups 2 ). Conversely, the statistical probability of a diagnosis increases with increasing concentration differences between the observed concentration and the cut-off concentration. In other words, diagnostic utility is reduced as concentrations converge towards the ‘diagnostic cut-off level’. It is therefore wrong to refer to these cut-off levels and concentrations close to them as being of diagnostic value when the reverse is true. This confusion arises from the unfortunate over-reliance on parametric statistics rather than using the more relevant but lesser used Bayes' theorem. 3 The latter adjust probability to concentration: intuitively higher probability is assigned as concentrations diverge away from the described cut-off level. It would be more appropriate to refer to these cut-off levels as the concentrations at which diagnosis cannot be made, i.e. ‘the non-diagnostic cut-off levels’.
DECLARATIONS